/Awesome-3D-Anomaly-Detection

Awesome-3D/Multimodal-Anomaly-Detection-and-Localization/Segmentation/3D-KD/3D-knowledge-distillation

MIT LicenseMIT

Awesome-3D-Anomaly-Detection-and-Localization Awesome

image image

- welcome to add if any information misses. 😎

Task Definition

Given a set of exclusively anomaly-free 3D scans of an object, the task is to detect and localize various types of anomalies

Dataset Intro

the first comprehensive dataset for unsupervised anomaly detection and localization in three-dimensional data. It consists of 4147 highresolution 3D point cloud scans from 10 real-world object categories. While the training and validation sets only contain anomaly-free data, the samples in the test set contain various types of anomalies. Precise ground truth annotations are provided for each anomaly.

Dataset Description

  • Five of the object categories in our dataset exhibit considerable natural variations from sample to sample. These are bagel, carrot, cookie, peach, and potato.
  • Three more objects, foam, rope, and tire, have a standardized appearance but can be easily deformed.
  • The two remaining objects, cable gland and dowel, are rigid.

2023

2022

image image

image

image

image

  • The MVTec 3D-AD Dataset for Unsupervised 3D Anomaly Detection and Localization [VISAPP 2022]

paper:https://arxiv.org/pdf/2112.09045.pdf

@misc{bergmann2021mvtec,
      title={The MVTec 3D-AD Dataset for Unsupervised 3D Anomaly Detection and Localization}, 
      author={Paul Bergmann and Xin Jin and David Sattlegger and Carsten Steger},
      year={2021},
      eprint={2112.09045},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}

Useful Links

Welcome to comments and discussions!!

Xiaohao Xu: xxh11102019@outlook.com

License

This project is released under the Mit license. See LICENSE for additional details.